Domain Walls, Black Holes, and Supersymmetric Quantum ...from the variation of the Ricci scalar....
Transcript of Domain Walls, Black Holes, and Supersymmetric Quantum ...from the variation of the Ricci scalar....
arX
iv:h
ep-t
h/01
0111
9 v3
12
Mar
200
1
hep-th/0101119ITEP-TH-71/00CALT 68-2306CITUSC/00-062HU-EP-00/54SLAC-PUB-8749
Domain Walls, Black Holes, and
Supersymmetric Quantum Mechanics
Klaus Behrndta1 , Sergei Gukovb2 and Marina Shmakovac3
a Institut fur Physik, Humboldt Universitat, 10115 Berlin, Germany
b Department of Physics, Caltech, Pasadena, CA 91125, USA
CIT-USC Center for Theoretical Physics, UCS, Los Angeles, CA 90089, USA
c CIPA, 366 Cambridge Avenue Palo Alto, CA 94306, USA
Stanford Linear Accelerator Center, Stanford University, Stanford, CA 94309, USA
Abstract
Supersymmetric solutions, such as BPS domain walls or black holes, in four- and five-
dimensional supergravity theories with eight supercharges can be described by effective
quantum mechanics with a potential term. We show how properties of the latter theory
can help us to learn about the physics of supersymmetric vacua and BPS solutions in these
supergravity theories. The general approach is illustrated in a number of specific examples
where scalar fields of matter multiplets take values in symmetric coset spaces.
January 2001
1 Email: [email protected] Email: [email protected] Email: [email protected]
*Work supported by DOE Contract DE-AC03-76SF00515.
1. Introduction and Summary
Supersymmetric vacuum configurations in gauged supergravity theories in four and
five dimensions recently receive a lot of attention due to their relevance to AdS/CFT
correspondence [1], and, more generally, to holography in non-superconformal field theories
and Domain Wall/QFT correspondence [2]. Of particular interest are BPS solutions which
preserve at least four supersymmetries.
In many of such supergravity solutions, the values of scalars and other fields depend
only on a single spatial coordinate. For example, in the case of a BPS black hole this is a
radial coordinate, while for a domain wall this is a transverse spatial coordinate. In any
case, one can integrate out the dynamics in the other directions which are isometries of
the solution and obtain a 0+1 dimensional theory – supersymmetric quantum mechan-
ics – where the distinguished spatial coordinate plays the role of time. This relation to
supersymmetric quantum mechanics was used in certain supergravity solutions, see e.g.
[3]. Interpretation of the attractor flow as supersymmetric quantum mechanics was also
suggested in [4].
In this work we treat both systems in a unified framework of the effective supersym-
metric quantum mechanics with a certain superpotential, so that critical points of the
superpotential correspond to supersymmetric (anti de Sitter) vacua in the supergravity
theory. In the case of domain walls the superpotential is inherited from the original su-
pergravity theory, and in the case of black holes it has the meaning of the central charge
of a black hole. Notice, that in both cases physics requires us to extremize these quanti-
ties. Although the presence of the superpotential is crucial, a lot of interesting questions
can be answered without precise knowledge of its form, simply assuming that it is generic
enough4.
For example, domain walls (and similarly black holes) interpolating between different
minima have interpretation of the RG-flow in the holographic dual field theory on the
boundary. Therefore, physically interesting questions about supersymmetric vacua and
RG-flow trajectories translate into classification of ground states and gradient flows in
the effective quantum mechanics. Following the seminal work of Witten [5], we use the
relation between supersymmetric quantum mechanics and Morse theory to classify these
supersymmetric vacuum configurations. Namely, according to Morse theory, the complex of
critical points with maps induced by gradient flow trajectories is equivalent to the Hodge-de
4 Later we will explain the precise meaning of these words.
1
Rham complex of the scalar field manifold. Notice, that the result is completely determined
by the topology of the scalar field manifold, but not the form of the superpotential.
With these goals and motivation, we start in the next section with the derivation
of supersymmetric quantum mechanics from five-dimensional BPS domain walls which
interpolate between different (AdS) vacua. In the dual four-dimensional theory these
solutions can be interpreted as RG trajectories in the space of coupling constants. We
show that these trajectories are gradient flows with a potential, so-called height function,
given by (the logarithm of) the superpotential in supergravity theory. We also explain
the relation between critical points of the height function and supersymmetric vacua of
the five-dimensional supergravity. In section 3 we extend these results to four-dimensional
gauged supergravity, and also include interaction with hypermultiplets which has not been
studied until recently. In section 4 we draw a parallel with the black hole physics and,
in particular, show how quantum mechanics of the same type appears from the radial
evolution. In this case, the height function is given by the central charge of the black hole.
We also derive effective potentials associated with membranes wrapped over holomorphic
curves in Calabi-Yau compactifications of M theory. After all these systems are reduced
to supersymmetric quantum mechanics, one might hope to achieve classification of the
critical points of the height function by means of Morse theory, cf. [5]. We review the
relevant topology and explain its physical interpretation in section 5. Finally, we put all
the ideas together in section 6 and demonstrate them in a family of simple examples based
on SL(3) symmetric coset spaces.
2. Quantum Mechanics of Domain Walls
BPS domain walls are kink solutions where the scalar fields interpolate between differ-
ent extrema of the supergravity potential and due to the AdS/CFT correspondence, these
solutions are expected to encode the RG flow of the dual field theory. Let us focus here on
the 5-d case with real scalars and postpone the modifications for complex or quaternionic
scalars for the next section. A Poincare-invariant ansatz for the metric reads as follows:
ds2 = e2U(− dt2 + d~x2
)+ dy2 . (2.1)
The function U = U(y) is fixed by the equations of motion coming from the variation of
the action:
S =
∫
M
(R2− 1
2gij∂φ
i∂φj − V)−∫
∂M
K . (2.2)
2
We included a surface term K as the outer curvature which cancels the surface contribution
from the variation of the Ricci scalar. There are no surface terms including the scalars
because they asymptotically extremize the superpotential and hence are constant at the
boundary. The form of the potential
V = 6( 3
4gij∂iW∂jW −W 2
)(2.3)
as a function of the superpotential W is universal for a given dimension and follows from
very general stability arguments [6]. In fact, for complex and quaternionic scalar field
manifolds it is also possible to define a real superpotential W . In this case the supergravity
potential V will have the similar form [7]; see the next section.
The Poincare invariance of the ansatz (2.1) implies that all worldvolume directions
are Abelian isometries, so that we can integrate them out. For our ansatz the Ricci scalar
takes the form R = −20(U)2 − 8U and after a Wick rotation to an Euclidean time we find
the resulting 1-dimensional action5:
S ∼∫
dy e4U[− 6 U2 +
1
2gij φ
iφj + V]. (2.4)
In deriving this expression, the surface term in (2.2) was canceled by the total derivative
term. The equations of motion of this action describe trajectories φi = φi(y) of particles
in the target spaceM with the metric gij . As a consequence of the 5-d Einstein equations,
these trajectories are subject to the constraint
−6 U2 +1
2|φi|2 − V = 0 (2.5)
with |φi|2 = gij∂yφi∂yφ
j. In order to derive the Bogomol’nyi bound we can insert the
potential into (2.4) and write the action as
S ∼∫
dy e4U[− 6 (U ∓W )2 +
1
2
∣∣φi ± 3 ∂iW∣∣2]∓ 3
∫dy
d
dy
[e4U W
](2.6)
leading to the BPS equations for the function U = U(y) and φi = φi(y):
U = ±W , φi = ∓3 gij∂W
∂φj. (2.7)
If these equations are satisfied, the bulk part of the action vanishes and only the surface
term contributes. In the asymptotically AdS5 vacuum this surface term diverges near
5 In our notation, dotted quantities always refer to y-derivatives.
3
the AdS boundary (U ∼ y → ∞) and after subtracting the divergent vacuum energy
one obtains the expected result that the energy (tension) of the wall is proportional to
∆W0 = W+∞ −W−∞.
Our metric ansatz (2.1) was motivated by Poincare invariance which is not spoiled by
a reparameterization of the radial coordinate. We have set gyy = 1, which is one possibility
to fix this residual symmetry. On the other hand, we can also use this symmetry to solve
the first BPS equation W dy = ±dU , i.e. take U as the new radial coordinate. In this
coordinate system the metric reads
ds2 = e2U(− dt2 + d~x2
)+dU2
W 2. (2.8)
Repeating the same steps as before we obtain the Bogomol’nyi equations for the scalars
−φi = gij∂j log |W |3 = gij∂jh (2.9)
which follow from the one-dimensional action
S ∼∫dy[|φi|2 + gij∂ih∂jh
]=
∫dy |φi + gij∂jh|2 + (surface term) (2.10)
where h = 3 log |W |. As before, the field equations are subject to the constraint 12 |φi|2 −
gij∂ih∂jh = 0 and the surface term yields the central charge. Supersymmetric vacua
are given by the extrema of h and the number and type of such vacua can possibly be
determined by using Morse theory where h is called the height function, see [5] and below.
If there are more than two smoothly connected extrema of h, we can build kink
solutions corresponding to domain walls in the 4- or 5-dimensional supergravity. Let us
summarize the different types of domain walls and discuss their implications for the RG
flow and Randall-Sundrum scenario, see also [8]. As long as W 6= 0 at the extremum, we
obtain an AdS vacuum and since the extrema of W are universal (independent of the radial
parameterization), we have to reach the AdS vacuum either near the boundary (U → +∞)
or near the Killing horizon (U → −∞). Obviously, the case U → +∞ corresponds to a large
supergravity length scale and therefore, due to the AdS/CFT correspondence, describes
the UV region of the dual field theory. The opposite happens for U → −∞, which is
related to small supergravity length scales and thus encodes the IR behavior of the dual
field theory. Moreover, extrema of W are fixed points of the scalar flow equations and
translate into fixed points of the RG flow, which can be either UV or IR attractive. The
4
universality of the fixed points of the RG flow, e.g. the scheme independence of scaling
dimensions, translates in supergravity to the fact that the properties of the extrema of the
superpotential are independent of the chosen parameterization of the scalar manifold.
In order to identify the different fixed points we do not need to solve the equations
explicitly; as we will see, they are determined by the eigenvalues of the Hessian of the
height function h. This data depends only on the local behavior of the superpotential near
the fixed point. Let us go back to the BPS equations (2.7) and expand these equations
around a given fixed point with ∂iW∣∣0
= 0 at φi = φi0. The superpotential becomes
W = W0 +1
2(∂i∂iW )0 δφ
iδφj ± . . .
with δφi = φi − φi0, and the cosmological constant (inverse AdS radius) is given by Λ =
−W 20 = −1/R2
AdS. Hence, the scalar flow equations can be approximated by6:
δφi = −(gij∂j∂kW )0 δφk . (2.11)
Next, we can diagonalize the constant matrix (gij∂j∂kW )0 and find
Ωik = (gij∂j∂kW )0 = W01
3∆(i) δik (2.12)
where we absorbed the inverse length dimension into W0. The dimensionless eigenvalues
∆(i) coincide with the eigenvalues of ∂i∂jh. According to the AdS/CFT correspondence
[1], these eigenvalues are the scaling dimensions of the corresponding perturbations in the
dual field theory. Namely, in a linearized version, the equations of motion for the scalars
become ∂2φi−M ijφj = 0 and the mass matrix readsM i
j = ∂i∂jV∣∣0
= W 20 ∆(i)(∆(i)−4)δij
or, measured in the units of W0, the mass formula becomes
(m(i))2 = ∆(i)(∆(i) − 4) . (2.13)
Consequently, near the AdS vacuum we find a solution to (2.11)
U = (y − y0)W0 , δφi = e−13 ∆(i)W0(y−y0) = e−
13 ∆(i)U . (2.14)
This approximate solution is, of course, valid only if δφi = φi−φi0 → 0 in the AdS vacuum
where U → ±∞ and therefore all eigenvalues ∆(i) have the same sign: ∆(i) > 0 for UV
6 For definiteness we took the upper sign convention.
5
fixed points (U → +∞), or ∆(i) < 0 for IR fixed points (U → −∞). Equivalently, UV
fixed points are minima of the height function h whereas IR fixed points are maxima. For
this conclusion we assumed that the scalar metric has Euclidean signature and W0 > 0. It
is important to notice that in the definition of the scaling dimensions the matrix Ωij has
one upper index and one lower index. It is straightforward to consider also the possibilities
W0 < 0 and/or timelike components of the scalar field metric. Note, the sign ambiguity in
the BPS equations (2.7) interchanges both sides of the wall, i.e. it is related to the parity
transformation y ↔ −y, which also flips the fermionic projector onto the opposite chirality.
The eigenvalues of Ωik of different signs mean that the extremum is IR-attractive
for some scalars and UV-attractive for the other and, therefore, is not stable (a saddle
point of h). Moreover, using these saddle points to connect two maxima/minima of h
would violate the proposed c-theorem for domain walls [9,10,11,12]. Namely, multiplying
eq. (2.9) by gikφk one obtains
−h = −φi∂ih = gij φiφj ≥ 0 . (2.15)
Therefore, along the flow, the height function h has to behave strictly monotonic and at
the extrema it corresponds to the central charge of the dual conformal field theory [13]:
cCFT ∼ R3ADS = 1/|W0|3 = e−h0 . Recall that in our sign convention larger values of the
radial parameter U correspond to the UV region and are minima of the height function h.
If we start with the UV point (U = +∞) and go towards lower values of U , the c-theorem
states that h has to increase, either towards an IR fixed point (maximum) or towards
a positive pole in h (W 2 → ∞), which is singular in supergravity and corresponds to
cCFT = 0. On the other hand, if we start from an IR fixed point (U = −∞) and go towards
larger values of U , due to the c-theorem h has to decrease, either towards a minimum (UV
fixed point) or towards a negative pole (W 2 → 0), which is not singular in supergravity
and corresponds to flat spacetime, cCFT → ∞. An example is the asymptotically flat
3-brane, where the height function parameterizes the radius of the sphere, which diverges
asymptotically (indicating decompactification) and runs towards a finite value near the
horizon which is IR attractive in our language.
In summary, there are the following distinct types of supergravity flows, which are
classified by the type of the extremum of the height function or superpotential. Depending
on the eigenvalues of the Hessian of h, the extrema can be IR attractive (negative eigenval-
ues), UV attractive (positive eigenvalues) or flat space (singular eigenvalues). Generalizing
6
the above discussion and allowing also possible sign changes in W , the following kink so-
lutions are possible (in analogy to the situation in four dimensions [8]):
(i) flat ↔ IR
(ii) IR ↔ IR
(iii) IR ↔ UV
(iv) UV ↔ UV (singular wall)
(v) UV ↔ singularity (W 2 =∞).
Note, there is no kink solution between a UV fixed point and flat space, because the c-
theorem requires a monotonic h-function and the UV point corresponds to a minimum of h,
whereas the flat space case is a negative pole. Moreover, if there are two fixed points of the
same type on each side of the wall, W necessarily has to change sign implying that the wall
is either singular (pole in W ) or one has to pass a zero of W . In addition, between equal
fixed points no flow is possible (that would violate the c-theorem) and therefore this case
describes a static configuration, where the scalars do not flow. This is also what we would
expect in field theory, where the RG-flows go always between different fixed points. Recall,
although a zero of W means a singularity in h, the domain wall solution can nevertheless
be smooth. In models with mass deformations (∆(i) = 2), type (v) walls appear generically
for models which can be embedded into maximal supersymmetric models, whereas models
allowing type (iv) walls typically can not be embedded into maximal supersymmetric
models7, an explicit example is discussed in [14].
In the Randall-Sundrum scenario one is interested in the case (ii), because in this case
the warp factor of the metric vanishes exponentially on each side of the wall.
3. Potentials from Gauged Supergravity
Extrema of the supergravity potential V are vacua of the theory, but not all extrema
correspond to stable vacua. Instead, one can show [6] that stable vacua are extrema of a
superpotential W which defines the supergravity potential in d dimensions to be
V =(d− 2)(d − 1)
2
( d− 2
d− 1gij∂iW∂jW −W 2
).
Once again, the corresponding scalar flow equations look like φi = −gij∂jh with the height
function
h = (d− 2) log |W |.7 Maximal supersymmetric models typically have only one UV extremum.
7
We will now derive the real superpotential W for different models. Depending on the num-
ber of unbroken supercharges, only special superpotentials can appear in supersymmetric
models. Our primary interest are supergravity duals of field theories with 4 unbroken
supercharges. Therefore the scalars on the supergravity side are part of vector, tensor or
hyper multiplets that parameterize the product space:
M =MV/T ×MH.
Known potentials are related to: (i) gauging isometries of M or (ii) gauging the global R-
symmetry. In the first case, the scalars and fermions become charged whereas in the second
case the scalars remain neutral. As a consequence of the gauging, the supersymmetry
variations are altered and flat space is, in general, not a consistent vacuum. If the potential
has an extremum, it is replaced by an AdS vacuum, which was not welcomed in the early
days of supergravity, but fits very well in the AdS/CFT correspondence and the holographic
RG flow picture. Let us discuss the different cases in more detail.
3.1. Gauged Supergravity in 5 Dimensions
Supergravity in five dimensions needs at least 8 supercharges, and scalar fields can
be part of vector-, tensor- or hypermultiplets. Each (abelian) vector multiplet contains a
U(1) vector Aiµ, a gaugino λi and a real scalar φi (i = 1, . . . , nV ). On the other hand, a
hypermultiplet includes two hyperinos ζu and four real scalars qu (u = 1, . . . , 4nH). Finally,
the gravity multiplet has besides the graviton, the gravitino ψAm and the graviphoton A0µ.
On the vector multiplet side, supersymmetry is a powerful tool in determining allowed
corrections. In fact, all couplings entering the Langrangian are fixed in terms of the cubic
form [15]
F =1
6CIJKX
IXJXK . (3.1)
In a Calabi-Yau compactification of M-theory, the fields XI (I = 0, . . . , nV ) are related
to the Kahler class moduli M I by a rescalling XI = MI
V1/3 with the Calabi-Yau volume
V = 16CIJKM
IMJMK and the constants CIJK are the topological intersection numbers
[16]. The scalar fields φa(XI ) parameterize the spaceMV defined by F = 1 and the gauge-
and scalar-couplings are given by
GIJ = −1
2(∂I∂JF )F=1, gij = (∂iX
I∂jXJGIJ )F=1. (3.2)
8
Much less is known on the hypermultiplet side. The four real scalars of each hyper-
multiplet are combined to a quaternion and parameterize a quaternionic mannifold MH.
From the geometrical point of view, the hypermultiplet sector in four and five dimensional
supergravity is the same, see [17] for further details. In any compactification from string
or M-theory there is at least one hypermultiplet – the so-called universal hypermultiplet
that contains Calabi-Yau volume V. This name is a little bit misleading since when nH > 1
there is no unique way to single out one direction on a general quaternionic manifold [18].
As long as we have only this single hypermultiplet, its geometry is given by the coset spaceSU(2,1)U(2) . But after including further hypermultiplets, quantum corrections will deform the
space in a way which is rarely known.
Now let us turn to potentials resulting from gauged isometries. Both manifolds,MV
and MH , have a number of isometries [19], which can be gauged [20,21]. However the
flow equations of the scalars are not sensible to gauged isometries ofMV ; it yields only an
additional “D-flatness” constraint8 [22,21]. More interesting is the gauging of isometries
ofMH ; see [23,7,24] for explicit examples. This is a quaternionic space, which implies the
existence of three complex structures and an associated triplet of Kahler forms Kx. The
holonomy group is SU(2)×Sp(nH) and the Kahler forms have to be covariantly constant
with respect to the SU(2) connection. The isometries are generated by a set of Killing
vectors kuI
qu → qu + kuI εI
and the required gauge covariant derivatives become dqu → dqu + kuIAI . In order to keep
supersymmetry, this gauging has to preserve the quaternionic structure, which means that
the Killing vector has to be tri-holomorphic (in analogy with the holomorphicity in N=1
supergravity). This is the case if we can express them in terms of a triplet of Killing
prepotentials P xI (with the SU(2) index x = 1, 2, 3)
Kxuvk
vI = −∇uP xI ≡ −∂uP xI − εxyzωyuP zI . (3.3)
Here ωyu are the SU(2) connections related to the Kahler forms by Kxuv = −∇[uω
xv]. They
can be also expressed in terms of the complex structures and the quaternionic metric as:
Kxuv = (Jx) ru hrv, and using
∑x(Jx) ru (Jx) vr = −3 δ vu we can write the Killing vectors as
kuI = −∑x huv(Jx) rv ∇rP xI . Next, introducing an SU(2)-valued superpotential
W BA ≡W x (iσx) BA with : W x = XIP xI = XI(φ)P xI (q) (3.4)
8 See also below for the similar situation in 4 dimensions.
9
(σx are the Pauli matrixes) the fermionic supersymmetry variations [21] become
δψAm =DmεA − i
3W AB Γmε
B ,
δλAi =− i
2
[Γm∂mφ
i εA − 2i gij∂jWAB εB
],
δζα =− i√2V Aαu
[Γm∂mq
u − 2huv(Jx) rv ∇rW x)]εA
(3.5)
where A,B are SU(2) indices and ∂i ≡ ∂∂φi
. We dropped the gauge field contributions,
since they are not important for flat domain wall solutions; they would be important e.g.
for a worldvolume geometry R × S3. For supersymmetric vacua, W BA has to become
extremal (∂iWx = ∇uW x = 0, for all x = 1, 2, 3). One can show [7], that the SU(2)
phase of W does not contribute to the scalar flow equations and is absorbed by the SU(2)
connection entering the covariant derivative Dm. Moreover, combining all scalars (φi, qu)
that parameterize the space M = MV ×MH with the metric gij = diag(gij , huv) one
recovers the BPS equation (2.7) with the real-valued superpotential
W 2 =∑
x
W xW x . (3.6)
There is one especially simple example, where the superpotential is only U(1) valued,
i.e. the Killing prepotential has only one, say, P 3I component. In this case the SU(2)
covariant derivative in (3.3) becomes a partial derivative and we have the freedom to shift
the Killing prepotential by any constant P 3I → P 3
I + αI . In fact, one can even set the
Killing prepotential to zero and keep only the constants αI , which are the analogs of the
FI-terms in field theory. As a consequence, the Killing vectors vanish as well as the charges
of the scalars. But still, we have a non-trivial potential giving a mass to all vector scalars.
In this case the superpotential becomes
W (3)(φi) = (αIXI)F=1 , (3.7)
which is manifestly real valued. Due to the constraint F = 1 this potential yields an AdS
vacuum, where generically all vector-scalars are fixed and the moduli space MV of vacua
is lifted except for a discrete set of extremal points of W (3). From the 5-d supergravity
perspective this model has been discussed in [25], and important for the RG-flow is the
property [15,26]
∂i∂jW(3) =
2
3gijW
(3) + Tijk∂kW (3) . (3.8)
10
This relation implies that all scaling dimensions, as defined in (2.12), are ∆(i) = +2 and
fulfill a sum rule [27]:∑
i ∆(i) = 2n, with n = dimMV . Therefore, the flow is generated by
mass deformations in the field theory and the positive sign indicates that all fixed points
are UV attractive; IR attractive critical points are excluded for this model [22,28]. This
model can be obtained from Calabi-Yau compactification of M-theory in the presence of
non-trivial G-fluxes parameterized by αI [23] and the superpotential can be written as
[29,30]
W (3) =
∫
CY
K ∧Gflux (3.9)
where K is the Kahler 2-form. However this compactification yields an un-stabilized
Calabi-Yau volume and, as long as we treat it as a dynamical field [31,30], this compacti-
fication does not give flat space or AdS vacua. Nevertheless this run-away problem can be
avoided by more general hypermultiplet gauges, see [7]. It would be interesting to derive
the general SU(2)-valued superpotential in the same way from M-theory. For a recent
study of W x (with W 3 = 0 but W 1 6= 0 and W 2 6= 0) see [32].
3.2. Gauged Supergravity in 4 Dimensions
Supergravity in 4 dimensions needs at least 4 supercharges and allows for more general
(holomorphic) superpotentials which are not related to gauged isometries. In the generic
case, these models have an AdS vacuum with a dual 3-d field theory with only 2 (unbroken)
supercharges; for domain wall solutions see [8].
However, if we again focus our attention on models with 8 supercharges, potentials
have to be related to gauged isometries. The main difference from the situation in 5 dimen-
sions concerns the vector multiplet side. Since vector fields in four dimensions have only
two on-shell degrees of freedom, each vector multiplet has to contain two scalars (in order
to complete the bosonic degrees of freedom). These two real scalars can be combined into
a complex scalar zi and supersymmetry requires that they parameterize a special Kahler
space, see [17] for a review. On the other hand, the scalars qu entering the hypermultiplets
parameterize again a quaternionic space and the gauging of the corresponding isometries
goes completely analogous to the case in 5 dimensions. This time, however, it is reason-
able to use the symplectic notation of special geometry and we will use the holomorphic
symplectic section V =(XI(z), FI (z)
)with the symplectic product defining the Kahler
potential:
e−K/2 = 〈V,V〉 = i(XIFI −XIF I). (3.10)
11
Then, we obtain a similar superpotential, cf. eq. (3.4):
W BA ≡W x (iσx) BA with W x = XI P xI = XI(z)P xI (q) . (3.11)
Notice, now XI = XI(z) is a complex field. In N=2 supergravity in 4 dimensions one
also introduces a symplectic vector Fµν = (F Iµν , GI µν) for the gauge fields. Both gauge
fields are related to each other and we took the freedom to transform all gauge fields to
F Iµν and therefore only the XI component of the section enters W . As in 5 dimensions we
can express the supersymmetry variations [33,17] in terms of the superpotential (3.11)
δψAm =DmεA − i
3eK/2 W A
B γmεB ,
δλAi =− i
2
[γm∂mz
i εA + 2i eK/2 gij∇jWA
B εB + kiIXIeK/2εA
],
δζα =− i√2V Aαu
[γm∂mq
u − 2 eK/2 huv(Jx) rv ∇rW x)]εA
(3.12)
with ∇jW x = P xI (q)(
∂∂zj + ∂K
∂zj
)XI(z) as a Kahler covariant derivative and ∇rW x denotes
the SU(2) covariant derivative, see (3.3). We included also a possible gauging of MV
related to the Killing vector kiI and remarkably this gauging affects only the gaugino
variation δλAi, but not the gravitino variation δψAm.
As usual the fermionic projector is derived from the timelike gravitino variation δψA0
and using the metric ansatz (2.8) this projector becomes
γU εA ± i W
AB
|W | εB = 0 (3.13)
and we define a real superpotential by
W 2 = eK |W |2 =∑
x
eKW xWx
=∑
x
eKXIXJP xI P
xJ (3.14)
(the hyper scalars qu are 4nH real fields so that the Killing prepotentials are real). Note,
the projector (3.13) contains both, the SU(2) phase as well as the U(1) phase related to
the complex fields XI and in order to ensure the vanishing of the radial gravitino variation
δψAU , both phases have to be absorbed into the SU(2) (resp. U(1)) Kahler connection.
This can impose further constraints.
12
Using the projector it is straightforward to show that the gaugino variation yields two
equations9: a D-flatness constraint kiIXI
= 0 and, after fixing the sign ambiguity in the
projector, the expected flow equation for the scalars zi:
zi = −gij∂zj logW 2 . (3.15)
Here we use: ∂iW2 = 2W∂iW =
∑x e
KW x(∂i + ∂iK)Wx
=∑
xW2
|W |2Wx∇iW
x. Finally,
following the steps done in 5 dimensions, see also [7], the hyperino variation yields the
same flow equation
qu = −huv∂v logW 2 .
As before, we can again consider the special case, where the Killing prepotentials have
only one component (e.g. P 1I = P 2
I = 0, P 3I 6= 0) and can be shifted by arbitrary constants
αI . The analog of (3.7) is now the complex superpotential [17]
W (3) = αIXI(z) .
Recall that this form is related to the special symplectic basis, where all GI µν gauge fields
have been dualized to F Iµν gauge fields. In general, the superpotential has the covariant
form [34,31]
W (3) = αIXI − βIFI . (3.16)
It allows for AdS vacua which are in one-to-one correspondence with solutions to the
attractor equations, which determine extrema of the supersymmetry central charge [35,36].
One can show that all these extrema are UV attractive [28] (due to a similar relation as
(3.8)) and, therefore, cannot give regular domain wall solutions. Like in 5 dimensions,
this superpotential is related to a compactification in the presence of fluxes. This time,
however, it is type IIB Calabi-Yau compactification and the superpotential can be written
as [31,29]
W (3) =
∫
CY
Ω ∧Hflux
where Ω is the holomorphic (3,0)-form and Hflux = HRRflux + τHNS
flux with τ as the com-
plexified scalar field of type IIB string theory. Supersymmetric vacua corresponding to the
9 In this matrix equation, the coefficient in front of each Pauli matrix has to vanish. In addition,
in all these calculations one has to keep in mind that the γ-matrix implicitly contains a W factor
due to its curved index.
13
minima of this superpotential have been recently studied in [32]. Moreover, the authors
of [32] pointed out a relation between these supersymmetric vacua and attractor points
of N = 2 supersymmetric black holes. Namely, they demonstrated the equivalence of the
supersymmetry conditions ∇iW x = 0 to the attractor equations that determine the values
of the scalar fields zi at the horizon of a black hole. We elaborate this relation in the next
section from the standpoint of the effective quantum mechanics, whereas in the rest of this
section we discuss ‘attractor’ interpretation of the other supersymmetry condition:
∇uW x = XI(∂uP
xI + εxyzωyuP
zI
)= 0 . (3.17)
The idea is that the supersymmetry variations of the hyperino (3.5) (or (3.12)) de-
scribe supersymmetric vacua of N = 2 gauged supergravity as well as BPS instantons (in
ungauged supergravity) responsible for non-perturbative corrections to the metric on the
moduli space of hypermultipletsMH . In order to see the relation more precisely, consider
supergravity theory obtained from Calabi-Yau compactification of M theory (or type IIA
string theory). Then, the instantons in questions are constructed from (bound states of
membranes and) Euclidean 5-branes wrapped on the Calabi-Yau manifold [37]. Put dif-
ferently, this BPS configuration can be understood as a five-brane wrapped around the
Calabi-Yau space with world-volume 3-form tensor field turned on.
Apart from the number of 5-branes, the instanton is characterized by the membrane
charge αu that takes values in the homology lattice H3(CY,ZZ). For a given set of the
charges, one can construct a spherically symmetric solution qu(r) which preserves half the
supersymmetry (3.5) and behaves like10 [37,38]:
qu(r) ∼ αu3r3
+ const, r →∞ . (3.18)
When r varies from r =∞ to r = 0, the scalar fields qu go to the fixed values determined
by the charges αu, similar to the attractor mechanism [35]. It was shown in [37], that the
radial evolution of the scalar fields qu(r) is described by null-geodesics in the effective 0+1
dimensional theory with target space MH. Up to boundary terms, we expect the action
of this effective theory is equivalent to the action of supersymmetric quantum mechanics
(2.10) with the superpotential (3.4) (or (3.11)). Thus, we claim that the flow on the moduli
space of the hypermultiplets is governed by W x.
10 For the sake of concretness, we consider five-dimensional case obtained from compactification
of M theory. Reduction to four dimensions is straightforward.
14
Since the instanton solution is expected to be smooth at r = 0, from the equation
(3.5) (with the U(1)R gauge field contribution included) we find that the fixed points of
the flow on the moduli space of hypermultiplets are characterized by the condition (3.17):
XIkuI = 0 . (3.19)
This simple condition means that the Killing vectors obey certain linear relations at the
fixed points in the space MH . Note, that in the hypermultiplet version of the attractor
mechanism we have fixed limit cycles rather than fixed points11. It would be interesting
to better understand the physics and the geometry of these fixed cycles in MH [39].
Intuitively, the hypermultiplet attractor equation (3.19) could be expected by analogy
with N = 2 BPS black hole solutions that exhibit enhancement of supersymmetry at the
horizon. For a similar reason, one might expect enhancement of supersymmetry at the
center, r = 0, of the BPS instanton constructed from a bound state of a membrane
wrapped around a special Lagrangian cycle and a five-brane wrapped on the entire Calabi-
Yau space. As we explained in the earlier sections, a supersymmetric vacuum is given by
the extremum of the superpotential (3.17) which, in turn, is equivalent to (3.19).
4. Black Hole – Domain Wall Correspondence
Similarity between physics of black holes of ungauged supergravity and the domain
wall solutions in gauged supergravity with gauging of vector multiplets [40,41] is based
on the fact that both theories can be described by the same one-dimensional effective
Lagrangian. In both cases the solution is fixed by the same set of attractor equations
[35,36] and the superpotentialW of gauged supergravity corresponds to the supersymmetry
central charge Z.
The D=4 N=2 action of ungauged supergravity in terms of special geometry has a
well known form:
L =R
2−Gij∂µzi∂µzj − huv∂µqu∂µqv − ImNΛΣFΛFΣ −ReNΛΣFΛ∗FΣ (4.1)
where z as in Section (3.2) are the complex scalars of vector multiplets parameterizing
special Kahler manifold MV with metric Gij = ∂2K
∂zi∂zjand Kahler potential K. Their
11 We would like to thank A. Strominger for pointing this out to us and drawing our attention
to the hypermultiplet version of the attractor mechanism as we describe it here.
15
kinetic term can be written in the form similar to the one in (2.2): gijdφidφj = Gij∂z
i∂zj .
The vector couplings ImN and ReN depend only on scalar fields z. Real scalars qu belong
to hypermultiplets and parameterize the quaternionic manifoldMH with the metric huv.
In ungauged supergravity the hypermultiplets are decoupled from the theory and we will
consider only vector multiplet scalars. We will choose the following ansatz for D=4 extreme
black hole metric [42]:
ds2 = −e2Udt2 + e−2U
(dτ 2
τ 4+
1
τ 2dΩ2
2
).
This assumption leads to one-dimensional effective action of the familiar form:
Seff ∼∫ ((
dU
dτ
)2
+ gijdφi
dτ
dφj
dτ+ e2UV (φ, p, q)
).
The potential V depends on the symplectic covariant electric and magnetic charges (pI , qI )
for symplectic vector Fµν = (F Iµν , GI µν) and can be identified with the symplectic invariant
form I1 in terms of the complex central charge (the graviphoton charge) Z = qIXI − pIFI
with the symplectic section (XI , FI):
V (φi, p, q) = I1 = |Z(z, p, q)|2 + |∇iZ(z, p, q)|2 (4.2)
where ∇i is a Kahler covariant derivative. Action (4.1) can be easily transformed to the
standard form of Bogomol’nyi bound:
L =
(dU
dτ± eU |Z|
)2
+∣∣dzdτ± eUGij∇jZ
∣∣2 ± d
dτ[eUZ]
with the equations of motion:
dU
dτ= ±eU |Z|, dzi
dτ= ±eUGij∇jZ. (4.3)
Using the same construction as in Section 3.2 for (3.15) we finally derived the equation
for the scalar field z:dzi
dU= Gij∂j logZ2.
This form is consistent with one-dimensional supersymmetric quantum mechanics.
16
For the black hole solutions in D=5 ungauged supergravity [36,42,38] the situation is
similar. The action of D=5 N=2 ungauged supergravity coupled to vector multiplets has
a form:
S5 =
∫R
2− 1
4GIJF
IF J − 1
2gij∂φ
i∂φj − e−1
48εµνρσλCIJKF
JµνF
JρσA
Kλ .
Scalar fields φi are defined through the constraint (3.1), F = 16CIJKX
IXJXK = 1, and
the gauge coupling GIJ and gij are given by (3.2) and depend only on XI and CIJK . The
ansatz for the D=5 black hole solutions is:
ds2 = −e−4Udt2 + e2U(dr2 + r2dΩ2
3
), (4.4)
and the ansatz vector fields is: GIJFItr = 1
4∂rKJ where KI = kI + QIr2 and QI are black
hole electric charges.
Using a new radial variable τ = 1r2 and equations of motion it is easy to reduce D=5
action to the effective D=1 dimensional action:
Seff ∼∫ (
3
(dU
dτ
)2
+1
2gij
dφi
dτ
dφj
dτ+ e4U 1
12V (φ, p, q)
)(4.5)
where the potential V comes from the kinetic term for the vector fields
V =3
2QIQJG
IJ = Z2 +3
2gij∂Zi∂Zj (4.6)
where Z = XIQI is a central charge. In (4.6) we used special geometry relations (see for
example [43]): gij = GIJXI,iX
J,j = −3CIJX
I,iX
J,j and gijXI
,iXJ,j = GIJ − 2
3XIXJ where
XI,i = ∂XI
∂φi .
The effective action (4.6) is easily changed to the familiar BPS form:
L = 3
(dU
dτ± 1
6e2UZ
)2
+1
2
∣∣dφi
dτ± 1
2e2U∂iZ
∣∣2 ∓ 1
2
d
dτ[e2UZ]
where |dφidτ |2 = gijdφi
dτdφj
dτ . Once again we have a system of linear differential equations of
the form:∂U
∂τ± 1
6e2UZ = 0, gij
∂φj
∂τ± 1
2e2U ∂Z
∂φi= 0.
Using the first equation to introduce the new radial variable dU = 16e
2Udτ we get the
familiar equation for the gradient flow in a scalar field manifoldMV :
∂φi
∂U= gij∂j logZ3.
17
This consideration shows that both cases – black holes of ungauged supergravity
coupled to vector multiplets and domain walls of gauged supergravity can be treated using
the same approach of one-dimensional supersymmetric quantum mechanics.
The first interesting question connected to the above discussion is the form of the
black hole potential (4.6) . The same potential appears as a result of M-theory compact-
ification on Calabi-Yau threefolds in the presence of non-trivial G-fluxes [23,30,21] and
corresponds to gauging of the universal hypermultiplet of D=5 dimensional theory. An
electrically charged five dimensional black hole corresponds to a membrane in M-theory
wrapped around 2-cycles of Calabi-Yau manifold in the process of compactification. We
can formally regard the M-theory compactification in the presence of the membrane-source
and corresponding flux.
The 11-dimensional supergravity theory is described by the action:
S11 =1
2
∫
M11
(√−gR− 1
2G ∧ ∗G− 1
6C ∧G ∧G
)
where C is a 3-form field with 4-form field strength G(4) = dC. In the presence of an
electrically charged membrane source Bianchi identities and equations of motion forD = 11
theory are:
dG = 0 (4.7)
d?G = d∗G+1
2G ∧G = 2k2
11(∗J) (4.8)
where ?G = ∂L∂G ,
∗ is Hodge-dual and J is a source current with the corresponding Noether
“electric” charge
Q =√
2k11
∫
M8
(∗J) =1√
2k11
∫
S7
(?G)(7)
where M8 is a volume and S7 is a sphere around the membrane. A formal solution to the
equations of motion (4.8) reads:
?G =∗ G+1
2C ∧ G =
√2k11Q
ε7Ω7
where ε7 is a volume form and Ω7 is a 7-volume. This solution corresponds to the vector
potential of the form:
C ∼ Q
r6ω3 (4.9)
where r is a D = 11 transverse distance to the membrane and ω3 is the membrane volume
form. Compactification from D = 11 dimensions down to D = 5 leads to the configuration
when the membrane is wrapped around compact dimensions.
18
A natural splitting of moduli coordinates Ma ⇒ (XI = MI
V1/3 ;V) and the condition
for function (3.1) : F (X) = 1 defines (h1,1 − 1)-dimensional hypersurface on the Calabi-
Yau cone and (h1,1 − 1) independent coordinates φa(XI ) (special coordinates) on this
hypersurface define vector multiplet moduli space with Kahler metric GIJ (3.2) :
GIJ (X) =i
2V
∫ωI ∧ ∗ωJ = −1
2∂I∂J logF (X)|F=1. (4.10)
A solution (4.9) of the equations of motion take the form:
G =1
V1
r3dt ∧ dr ∧ αIωI . (4.11)
Here r is a D = 5 transverse radial coordinate and V is a Clabi-Yau volume, and “electric”
charges αI = GIJαJ are
αI =
∫
C(4)I×S3
(∗G)
where C(4)I, I = 1, ..., h1,1 are 4-cycles in the Calabi-Yau manifold. The nontrivial flux
of this form leads to the appearance of the nonzero scalar potential. From the G∧∗G term
of the action it follows that∫
M11
G ∧∗ G =
∫
M11
√gstr
r6
1
V2αIαJωI ∧ ∗ωJ = 2
∫
M5
√gE
r6
1
V2αIαJG
IJ(X) (4.12)
here we use the relation√gstr = V−5
3√gE . The effective potential for D = 5 is:
V5 =1
r6
1
V2αIαJG
IJ(X).
This D=5 potential V5 explicitly depends on r and fells with the distance to the source
so that in this case there is no run-away Calabi-Yau volume as in [23,30].
In the effective one-dimensional theory (4.5) U plays a role of a dilaton and the
appearence of the effective potential (4.6) can be understood as a result of gauging of an
additional axionic shift symmetry.
The consideration of Section 3 shows that hypermultiplets can also play a very impor-
tant role in the physics of black holes. The difference between a black hole solution and a
domain wall solution is in the presence of non-trivial vector fields in the black hole case.
Those fields should be added to the supersymmetry variations (3.5) (see [21]):
δψAm =DmεA +
i
8XI(Γ
npm − 4δnmΓp)F Inpε
A − i
3W AB Γmε
B ,
δλAi =− i
2
[Γm∂mφ
i εA − 2i gij∂jWAB εB
]+
3
8gij∂jXIΓ
mnF ImnεA ,
δζα =− i√2V Aαu
[Γm∂mq
u − 2huv(Jx) rv ∇rW x)]εA.
(4.13)
19
It is easy to see that in ungauged supergravity hypermultiplets are decoupled from
the theory and do not affect the solution. It may not be the case when some gauging is
present.
An application of Morse theory to the black hole physics can have interesting conse-
quences. It is possible to consider a black hole solution in gauged supergravity with vector
multiplet gauging [44] . In this case hypermultiplets are also decoupled and, unfortunately
(as we will see later, rather, predictably), the BPS solutions contain naked singularities
or blow up near the horizon. This means that in this theory it is impossible to find a
black hole with a regular horizon embedded in the AdS5. On the other hand, application
of Morse theory to this case predicts existence of only one critical point (or several of the
same type) and the absence of a second non-trivial vacuum (see Section 5) is justified.
The case of the black hole solution with hypermultiplet gauging is much more com-
plicated. In this case, hypermultiplets are no longer decoupled from the theory and the
hypermultiplet gauging can lead to the appearance of a second non-trivial critical point.
It will be interesting to consider such solutions in the future work.
5. Morse Theory and Vacuum Degeneracy
In the previous sections we studied the effective dynamics of BPS solutions in N = 2
five-dimensional supergravity which preserve the SO(3, 1) (or SO(4)) subgroup of the
Lorentz symmetry, as well as the effective dynamics of the similar solutions in four dimen-
sions. The examples — such as supersymmetric domain walls or spherically symmetric
static black holes — correspond to solutions where space-time metric is a function of a
single coordinate. Effective dynamics of a BPS state with these properties turns out to
be supersymmetric quantum mechanics of the form (2.10), cf. [45,46,47]. The nature of
the BPS state is perfectly indifferent. For all such BPS states it is true that the solution
represents a gradient flow of the height function h between two critical points in the scalar
field manifoldM. For example, in five-dimensional gauged supergravity coupled to a cer-
tain number of vector multiplets M = MV is a Riemannian manifold parameterized by
scalar fields φA from the vector multiplets and h is (logarithm of) the superpotential W .
Moreover, if this theory is obtained from compactification of M theory on a Calabi-Yau
three-fold, thenMV is just the Kahler structure moduli space of this Calabi-Yau manifold,
and h has a microscopic interpretation in terms of a G-flux [23,30].
20
In any case, the study of BPS objects described above and the classification of su-
persymmetric vacua in D = 4 and D = 5 supergravity boils down to a simpler problem
in supersymmetric quantum mechanics with the effective action (2.10). This connection
to supersymmetric quantum mechanics allows one to address many interesting physical
questions studying more elementary system. For example, Klebanov and Tseytlin used
this relation to study supergravity duals of the RG-flows in SU(N) × SU(N +M) gauge
theories [3]. In this section we will discuss another application of this relation.
For a compact space M, it has been shown by Witten [5] that supersymmetric quan-
tum mechanics with the action (2.10) is just Hodge-de Rham theory of M. Therefore,
assuming it is also the case for certain (non-compact) scalar field manifolds M that oc-
cur in supergravity, we can use topological methods — in particular, Morse theory — to
classify possible BPS states and supersymmetric vacua they connect. As we explained
above, our results are quite generic irregardless of the physical nature of a given BPS
solution; they work equally well for BPS domain walls, for spherically symmetric black
holes, or any other BPS object with the effective action (2.10). We only make a couple of
assumptions necessary in the following. First, we requireM to be a Riemannian manifold,
although sometimes it comes with some additional structure (e.g. complex or quaternionic
structure) corresponding to additional (super-)symmetry in the problem. One has to be
very careful with this assumption in the cases when M has singularities, in particular, in
Calabi-Yau compactifications of M theory. Second, we assume that the space M is either
compact or it has the right behaviour at infinity, so that the topological methods of Morse
theory are reliable12.
Given a Riemannian manifoldM, let h:M→ R be a ”good” Morse function of C∞-
class. This condition means that every point in M is either a regular point where dh 6= 0
or an isolated critical point p ∈ M where dh = 0 and h can be written as:
h(φA) = h(p) −k∑
A=1
(φA)2 +n∑
A=k+1
(φA)2 (5.1)
in some neighborhood of p.
The number k in (5.1) is called the Morse index of a critical point p and is denoted
by µ(p). It turns out that µ(p) has a nice physical interpretation in supergravity. Recall,
12 For an analog of Morse theory in complex non-compact geometry that admits a holomorphic
torus action see [48].
21
that the critical points of h correspond to supersymmetric vacua. Moreover, according to
the analysis of section 2, the eigenvalues of the Hessian of h determine the type of the
attractor point. In particular, critical points with zero Morse index are UV attractive. On
the other hand, critical points with µ = n are IR attractive. In general, a critical point
p is of mixed type, 0 < µ < n. In models with only vector multiplets, the superpotential
depends generically on all scalars and such mixed critical points do not correspond to
stable vacua, see discussion after (2.14).
Therefore, in order to classify possible supersymmetric vacua, one has to know how
many points have a given Morse index. Morse theory provides a nice answer to this problem
in terms of the topology of the scalar field manifold M. Before we state the result let us
introduce a few more notions which will be convenient in the following.
As we explained in the previous sections, a supersymmetric domain wall or a spheri-
cally symmetric black hole corresponds to a gradient flow13 of h from one critical point p
to another critical point q. Consider the “moduli space of these BPS solutions” M(p, q).
Namely, for a pair of critical points p and q we define:
M(p, q) = φ: R→M | dφdt
= −∇h, limt→−∞
φ(t) = p, limt→+∞
φ(t) = q/ ∼
to be the moduli space of the gradient trajectories from p to q modulo the equivalence
relation φ(t) ∼ φ(t+ const). In general, M(p, q) is not a manifold, though perturbing it a
bit we can always assume that it is a manifold (or a collection of points, if we have a finite
number of distinct gradient trajectories).
A classical result in Morse theory asserts that the real dimension of M(p, q) is given
by:
dimM(p, q) = µ(p) − µ(q) − 1. (5.2)
It immediately follows that non-singular BPS domain walls (or spherically symmetric black
holes) of types (ii) and (iv) interpolating between two vacua of the same kind do not exist
in theories where h is a global Morse function, cf. the discussion in the end of section 2.
Indeed, if there are two critical points of the same type (i.e. either both IR or both UV),
then the virtual dimension (5.2) becomes negative. Another physical result that follows
from (5.2) is that domain walls of type (iii) connecting an IR attractive point and a UV
attractive point come in (n− 1)-dimensional families.
13 It is this place where we use the Riemannian metric on M to define a gradient vector field
∇h.
22
It is curious to note that BPS solutions interpolating between mixed (IR/UV) vacua
also appear in Morse theory as boundary components ofM(p, q). For example, codimension
1 boundary of M(p, q) consists of the points corresponding to two consequent gradient
flows: first from p to some other critical point s, µ(p) > µ(s) > µ(q), and then from s to q:
∂M(p, q) = ∪s M(p, s) ×M(s, q).
Similarly, codimension 2 boundary consists of the points corresponding to three consequent
flows p → s → r → q with µ(p) > µ(s) > µ(r) > µ(q), etc. Including all these boundary
components we obtain a compact oriented manifold M (p, q).
Now we define Witten complex C∗(M, h) as a free abelian group generated by the set
of critical points of h:
Ck(M, h) = ⊕µ(p)=kZZ · [p].
Furthermore, we define ∂:Ck(M, h)→ Ck−1(M, h) via the sum over gradient trajectories
(counted with signs):
∂[p] =∑
µ(q)=k−1
#M(p, q) [q].
Then, the Morse-Thom-Smale-Witten theorem says that ∂2 = 0 and:
H∗(C∗(M, h)) = H∗(M, ZZ). (5.3)
Therefore, the classification of critical points and gradient trajectories which represent
supersymmetric vacua and BPS solutions, respectively, can be addressed in terms of the
topology of M. In particular, one finds the following lower bound on the total number of
supersymmetric vacua:
∑rank Ci(M, h) ≥
∑rank Hi(M,R). (5.4)
This is the classical Morse inequality.
We conclude that by using Morse theory formulas (5.2) – (5.4) one can classify five-
dimensional systems discussed in sections 2 and 3. In what follows we illustrate these
methods in a number of interesting examples and, in particular, we count the number
of supersymmetric vacua computing the Betti numbers bi = rank Hi(M,R) of the cor-
responding scalar field manifolds. Even though the general formulas (5.2) – (5.4) were
derived for a compact manifold M, the Morse inequality (5.4) is expected to hold in a
23
wider class of examples, including certain non-compact manifolds MV and MH relevant
for supergravity. For example, let us consider N ≥ 2 five-dimensional gauged supergravity
interacting with a certain number of vector multiplets. As we will see in a moment, such
theories possess at most one supersymmetric critical point on every branch of the scalar
field manifoldMV , as long as interaction with other matter fields (e.g. hyper multiplets)
can be consistently ignored, and as long as the superpotential W is generic enough to be
considered as a good global Morse function.
First, consider examples with exactly N = 2 supersymmetry where scalar fields φA
take values in a real homogeneous cubic hypersurfaceMV = F = 1 defined by (3.1) in a
vector space parameterized by the fields XI with Minkowski signature [25]. We expect that
real cubics of this form have trivial topology. Even though we do not have a mathematical
proof of this fact, we argue as follows. Suppose, on the contrary, that MV is topologically
non-trivial. Then, there should exist at least two critical points, one of which must have
non-zero Morse index, 0 < µ < nV . However, as we explained in section 2, the existence
of such points would violate the c-theorem [11] and contradict our original assumtion that
MV is a Riemannian manifold with a positive-definite metric.
Now let’s see that scalar field manifoldMV in N > 2 five-dimensional gauged super-
gravity is also topologically trivial. This result immediately follows from the fact14 that in
supergravity theories with more supersymmetry the spaceMV can always be represented
as a quotient space of a non-compact group G by its maximal compact subgroup H [42].
Note, that G can have time-like directions which would be inconsistent if we did not divide
by H that makes the metric on the quotient spaceMV = G/H positive-definite. Another
effect of the quotient by H is that the space G/H is topologically trivial unless we divide
further by a discrete symmetry group, e.g. U-duality group15. This is one way to make
the topology of M topologically non-trivial.
The second possibility to get theories with multiple supersymmetric vacua is to take
a space of scalar fields M with more than one branch. For instance, an important family
of theories is based on the target spaces of the following simple form:
MV = SO(n, 1)/SO(n) : Since SO(n) is the maximal compact subgroup in the non-
compact group SO(n, 1) the quotient space MV is equivalent, up to homotopy, to a set
14 We thank R. Kallosh and N. Warner for explaining this to us.15 We are grateful to E. Witten for pointing out that a quotient by a discrete group may lead
to interesting non-trivial topology of MV .
24
two points. The number of points comes from the number of disconnected components
in the non-compact group SO(n, 1) which looks like a hyperbolic space. Therefore, ir-
regardless of the value of n in this class of models we find the following Betti numbers
bi = rank Hi(MV ,R):
b0 = 2,
bi = 0, i > 0.
From the Morse inequality (5.4) it follows that the corresponding supergravity theories
with generic W possess at least two UV attractive supersymmetric vacua. By numerical
computations, one can verify that there are exactly two vacua of UV type. We remark that
there are no smooth domain walls because the two vacua belong to different disconnected
branches.
MV = SO(n − 1, 1) × SO(1, 1)/SO(n− 1) : Once again, in this example we divide
a non-compact group by its maximal compact subgroup, so that the resulting space is
isomorphic to a set of points for the reason explained above. This time we get 4 points,
one for every disconnected component of SO(n − 1, 1)× SO(1, 1). For the Betti numbers
of MV we obtain:
b0 = 4,
bi = 0, i > 0.
Similar to the previous example, there are at least four UV attractive vacua which can not
be smoothly connected by BPS domain walls.
These are examples of symmetric spaces which typically appear as scalar field mani-
folds in N = 2 five-dimensional supergravity theories and also in models with additional
(super-)symmetry structure [49], as we mentioned earlier.
It is important to stress here that it is crucial for the height function h to be globally
defined over the entire target space M. For example, this assumption breaks down in a
very important class of models corresponding to M theory compactification on Calabi-Yau
three-folds. In these models MV is just the Kahler structure moduli space of the Calabi-
Yau manifold. Since the Kahler structure moduli spaces usually have trivial topology, one
might naively conclude from (5.3) that there is only one vacuum (a UV attractive fixed
point) and no non-trivial domain walls. However, in general,MV consists of several Kahler
cones separated by the walls where certain algebraic curves in the Calabi-Yau space shrink
to zero size. Local anomaly arguments in heterotic M theory suggest that G-flux should
25
jump while crossing a Kahler wall [50]. In fact, passing through a flop transition point the
second Chern class changes and, therefore, the 5-brane charge induced in the boundary
field theory also jumps. Since the total 5-brane charge should be conserved (and equal to
zero in the compact Horava-Witten setup) some αI also have to jump, as if the flop curves
effectively carry a magnetic charge [50]. Although the superpotential (3.9) passes these
curves smoothly, because the corresponding XI vanishes there, its (second) derivatives
jump once we cross a Kahler wall, and the corresponding h is not a globally defined height
function.
A similar result occurs in gauged supergravity theories coupled to hyper multiplets.
In this case supersymmetry implies that scalar fields from hyper multiplets parameterize
a quaternionic manifoldMH of negative curvature [51]:
R = −8(n2H + 2nH)
Even though this condition is not as much restrictive as the supersymmetry condition
(3.8) in the case of vector multiplets, known examples of quaternionic homogeneous coset
spaces that may serve as hypermultiplet target manifolds are typically non-compact and
topologically trivial:
MH = SO(n, 4)/SO(n) × SO(4) : Similar to the exampleM = SO(n, 1)/SO(n), this
quaternionic space is contractible since16 SO(n)×SO(4) is the maximal compact subgroup
of SO(n, 4). Therefore, MH is homologous to a set of two points:
b0 = 2,
bi = 0, i > 0.
The same result we find forMH = SU(n, 2)/SU(n)×SU(2). As we mentioned earlier,
a simple way to obtain models where scalar fields take values in a topologically non-trivial
manifold MH is to devide by a discrete group which, for example, may be a subgroup of
the isometries of MH . For example, if we have only the universal hypermultiplet, n = 1,
the coset space MH = SU(1, 2)/U(1) × SU(2) is a quaternionic Kahler manifold, where
the Kahler potential can be written as K(S,C) = − log(S + S − 2(C + C)2). It has two
abelian isometries corresponding to shifts of the complex scalar fields S and C, S → S+ ia
16 Stricty speaking, S(O(n)×O(4)) is the maximal compact subgroup of SO(n, 4).
26
and C → C + ib. To get a topologically non-trivial manifold, we can consider a quotient
space:
MH/ZZ2 =SU(1, 2)
U(1) × SU(2)× ZZ2
where the action of ZZ2 is equivalent to identification S ∼ S+ ia and C ∼ C+ ib for integer
numbers a and b. Supergravity theory coupled to a hypermultiplet based on the resulting
quotient space is expected to have at least∑
i bi(MH/ZZ2) = 4 supersymmetric vacua.
Before we conclude this section, let us remark that due to its construction, the super-
potential or height function may not depend on all hyper-scalars and the chosen gauged
isometry of MH determines the scalars on which the superpotential depends. Hence,
for a given superpotential, obtained by a specific gauging, a mixed critical point with
0 < µ < nH may appear as a “good” UV or IR fixed point. Only the critical points
with µ = 0 and µ = nH are “gauge-independent” and appear in all gaugings as UV and
IR fixed points. Assuming that Morse inequalities are saturated, every component of Mhas exactly one UV critical point with µ = 0 and at most one IR point with µ = nH if
M is a compact manifold17. The values of all scalar fields are fixed in these two critical
points. Additional (mixed) critical points can be stable under the UV/IR scaling only if
the superpotential has (bad) flat directions.
6. Examples: SL(3) Symmetric Coset Spaces
Let us consider in more detail a specific example where the quotient space which
appears in N = 2 five-dimensional supergravity is associated with Jordan algebras of the
form:
M =Str0(J)
Aut(J)
where Str0(J) is the reduced structure group and Aut(J) is the automorphism group of
a real unital Jordan algebra of degree 3. A simple example based on irreducible J is
M = SL(3,R)/SO(3), which is a three-dimensional analog of the Lobachewsky plane,
SL(2,R)/SO(2). Like its two-dimensional analog, the space M is contractible because
a semisimple Lie group SL(n,R) is isomorphic to its maximal compact subgroup SO(n).
After we divide by the latter we get a point, up to homotopy:
b0 = 1 (6.1)
17 It should be stressed, however, that the authors do not know whether supergravity theories
based on compact scalar field manifolds M exist or not.
27
bi = 0, i > 0.
So, we come to the conclusion that the STU model with generic superpotential has a single
supersymmetric vacuum. In the rest of this section our goal will be to identify the physics
of this vacuum.
There are two generalizations of this example over complex numbers and quaternions.
In any case we divide SL(3, F ), where F = R,C,H is the base field in question, by
its maximal compact subgroup. Although the resulting quotient space has trivial topol-
ogy (6.1) for all ground fields F , the physics is different. Namely, we claim that three
supersymmetric vacua corresponding to various choices of F describe AdS5 dual of N = 4
super-Yang-Mills perturbed by mass terms which preserve different subgroups of SO(6)
R-symmetry.
In order to see that a scalar vev. gives rise to a mass deformation, one has to find
scaling dimension ∆(i) of the corresponding operator O(i) in the boundary theory. Using
the standard formula (2.13) we get ∆(i) = 2 which allows us to identify O(i) with a mass
term. In general, a mass term is specified by a bosonic symmetric 6 × 6 matrix and a
fermionic symmetric 4 × 4 matrix. In our three exampes, however, these matrices have
additional symmetries reminiscent of N = 4 super-Yang-Mills theory with R-symmetry
twists [52]. In particular, at special values of twists, where the corresponding phases are
equal to (−1), extra hypermultiplets become massless18, and we expect to get a four-
dimensional superconformal theory 19.
Below we discuss in more details the case where F = C. In particular, we solve the flow
equations (2.7) for the coset space SL(3,C)/SU(3). This coset manifold is parameterized
by 8 non-trivial scalar fields with the intersection form [15]
F = STU − S|X|2 − T |Y |2 − U |Z|2 + 2 Re(XY Z) (6.2)
18 To see this, it is convenient to think of gauge theories with R-symmetry twists as compact-
ifications of five-dimensional gauge theories on a circle with twisted boundary conditions on S1
[52]. The masses of Kaluza-Klein states are given by m = n+ 1/2 + qW where W is the value of
the Wilson lines of the gauge field and q is the charge of a given mode. Note that for any n ∈ ZZ
we get a whole hyper-multiplet. Moreover, for n = 0 and −1 and W = 1/2 and q = +1 and −1
we find two massless hypermultiplets.19 We thank Ori Ganor for discussions on this point.
28
where X,Y and Z are complex and S, T and U are real. The unique solution of the
attractor equations [36]: ∂IF = e−2U HI , which solves the flow equations (2.7) (see [30])
is given
S = (HTHU −1
4|HX |2) e−4U , X =
1
4(HYHZ − 2HXHS) e−4U ,
T = (HSHU −1
4|HY |2) e−4U , Y =
1
4(HXHZ − 2HYHT ) e−4U ,
U = (HSHT −1
4|HZ |2) e−4U , Z =
1
4(HXHY − 2HZHU ) e−4U
(6.3)
where HI is a set of harmonic functions
HI = hI + 6αI y
which are real for the S, T,U components and complex for the X,Y,Z components. For
the metric we take the ansatz:
ds2 = e2U[− dt2 + d~x2
]+ e−4Udy2 (6.4)
where the function e−2U is obtained from the requirement F = 1:
e6U = HSHTHU −1
4
(HS |HX |2 +HT |HY |2 +HU |HZ |2
)+
1
4Re(HXHYHZ) . (6.5)
The scalar fields are defined by F = 1 and we may consider, for example, the ratios:
φA = TS , US , XS , YS , ZS . The uniqueness of the solution fits very well with our expectation
from Morse theory, that this coset allows only a unique critical point. At the critical point
the space time becomes AdS5 as y → +∞ with the negative cosmological constant:
Λ = −(e4U/y2)y→+∞
In the case of a unique critical point with a negative cosmological constant, it is natural
to ask what is the four-dimensional field theory dual to this AdS5 vacuum.
Notice that the scalar fields φA defined as ratios of harmonic functions stay finite in
the AdS vacuum and approach their critical values. Having the explicit solution, we can
also calculate the supergravity effective action. As given by (2.6), the BPS nature ensures
vanishing of the bulk term and the surface part yields20:
Seff =2
3
(∂ye
6U)y=+∞
=(− 2Λy2 + a1 y
)y→∞
+ a2. (6.6)
20 Note the different coordinate system.
29
As expected, this action has singular terms and a finite part. The leading singularity scales
with the cosmological constant (AdS volume) while the subleading term a1 and the finite
term a2 can be obtained by inserting the harmonic functions (6.5). The divergent part will
be subtracted by the renormalization in the field theory and the finite part will give the
renormalized effective action; in our approximation we see only potential or mass terms.
From the RG-flow point of view this AdS vacuum corresponds to an UV fixed point.
While moving towards negative y the warp factor e2U decreases monotonically (according
to c-theorem) and we approach the IR region in the fields theory. Because e6U is negative
at y = −∞, we have to pass a zero at some finite value of y, which is the singular end-point
of the RG flow. Like in any other case with vector multiplets only, the absence of an IR
fixed point forces the solution to run into a singularity.
On the other hand, from the string theory perspective this solution corresponds to a
5-brane wrapping a holomorphic 2-cycle, namely a torus T 2. Once again, we point out
the analogy with the construction of gauge theories with R-symmetry twists from (2,0)
theory on a torus in the limit where the size of the torus and the values of twists tend to
zero (Vol(T 2) → 0 and α → 0) while their ratio α/Vol(T 2) remains finite and defines a
mass scale in the resulting theory [52]. From this construction it is clear that at the special
values of twists α, where the theory becomes superconformal, it must be dual to AdS5×S5
perturbed by a dimension-8 operator (proportional to Vol(T 2)) and dimension-2 operators
(proportional to α), similar to the AdS5 vacuum of SL(3) coset spaces we found.
It is reasonable to put another (5-brane) source at some place where the warp factor
is still positive, say y = 0. This extra source appears as a non-trivial right-hand side of
the harmonic equations:
∂2HI ∼ αIδ(y)
where αI is the component of the 5-brane charge related to a basis of (1, 1)-forms ωI .
There are two possibilities: in the first option we continue in a symmetric way through the
source, which implies the replacement HI → hI +6αI |y| and is equivalent to a sign change
in the flux vector αI while passing the source at y = 0. This case appears in the ZZ2 orbifold
of the Horava-Witten setup compactified to 5 dimensions [23]. But we may also consider
the case where the flux jumps from zero to a finite value, i.e. on the side behind the source
we can set αI = 0, which is equivalent to the replacement H → hI + 6αI12(y+ |y|), so that
HI is constant for negative y and the space time is flat.
By adding this source, we cut off the (singular) part of space-time and glue instead
an identical piece (Z2 symmetric) or flat space (vanishing flux on one side). If one wants
30
to discuss a RS-type scenario, one may also cut off the regular AdS part and keep on both
sides the naked singularity. In the first case the source is the standard positive tension
brane generating an AdS space on both sides, whereas in the second case a negative tension
brane is accompanied with a naked singularity. For a more detailed discussion of sources
see [53].
Acknowledgments
We are grateful Alexander Chervov, Ori Ganor, Brian Greene, Renata Kallosh, Eric
Sharpe, Gary Shiu, Andrew Strominger, Nicholas Warner, and Edward Witten for help-
ful discussions and comments. The work of K.B. was partly done at the Theory group
of Caltech and is supported by a Heisenberg grant of the DFG and by the European
Commission RTN programme HPRN-CT-2000-00131. S.G. is supported in part by the
Caltech Discovery Fund, grant RFBR No 98-01-00327 and Russian President’s grant No
00-15-99296.
31
References
[1] J.M. Maldacena, ”The Large N Limit of Superconformal Field Theories and Su-
pergravity”, Adv. Theor. Math. Phys. 2 (1998) 231; S.S. Gubser, I. R. Klebanov,
A. M. Polyakov, ”Gauge Theory Correlators from Non-Critical String Theory”, Phys.
Lett. B428 (1998) 105; E. Witten, ”Anti De Sitter Space And Holography”, Adv.
Theor. Math. Phys. 2 (1998) 253.
[2] H.J. Boonstra, K. Skenderis, P.K. Townsend, “The domain-wall/QFT correspon-
dence”, JHEP 9901 (1999) 003.
[3] I.R. Klebanov, A.A. Tseytlin, “Gravity Duals of Supersymmetric SU(N) x SU(N+M)
Gauge Theories,” Nucl.Phys. B578 (2000) 123.
[4] G. Moore, “Arithmetic and Attractors,” hep-th/9807087.
[5] E. Witten, “Supersymmetry and Morse theory”, J.Diff.Geom. 17(1982) 661.
[6] P.K. Townsend, “Positive energy and the scalar potential in higher dimensional (super)
gravity”, Phys. Lett. B148, 55 (1984).
[7] K. Behrndt, C. Herrmann, J. Louis and S. Thomas, “Domain walls in five dimensional
supergravity with non-trivial hypermultiplets,” hep-th/0008112.
[8] M. Cvetic and S. Griffies, “Domain walls in N=1 supergravity”, hep-th/9209117;
M. Cvetic and H. H. Soleng, “Supergravity domain walls,” Phys. Rept. 282 (1997)
159, [hep-th/9604090].
[9] E. Alvarez and C. Gomez, “Geometric holography, the renormalization group and the
c-theorem,” Nucl. Phys. B541 (1999) 441, hep-th/9807226.
[10] L. Girardello, M. Petrini, M. Porrati and A. Zaffaroni, “Novel local CFT and exact
results on perturbations of N = 4 super Yang-Mills from AdS dynamics,” JHEP 9812
(1998) 022 hep-th/9810126.
[11] D.Z. Freedman, S.S. Gubser, K. Pilch and N.P. Warner, “Renormalization group flow
from holography, supersymmetry and a c-theorem”, hep-th/9904017.
[12] K. Behrndt, “Domain walls of D = 5 supergravity and fixed points of N = 1 super
Yang-Mills,” Nucl. Phys. B573 (2000) 127, hep-th/9907070.
[13] M. Henningson and K. Skenderis, “Holography and the Weyl anomaly”JHEP 9807,
023 (1998), hep-th/9812032.
[14] K. Behrndt and M. Cvetic, “Supersymmetric domain wall world from D = 5 simple
gauged supergravity,” Phys. Lett. B475, 253 (2000), hep-th/9909058.
[15] M. Gunaydin, G. Sierra and P. K. Townsend, “The geometry of N=2 Maxwell-Einstein
supergravity and Jordan algebras”, Nucl. Phys. B242, 244 (1984).
[16] A. C. Cadavid, A. Ceresole, R. D’Auria and S. Ferrara, “Eleven-dimensional super-
gravity compactified on Calabi-Yau threefolds,” Phys. Lett. B357, 76 (1995), hep-
th/9506144.
32
[17] L. Andrianopoli, M. Bertolini, A. Ceresole, R. D’Auria, S. Ferrara, P. Fre and T. Magri,
“N = 2 supergravity and N = 2 super Yang-Mills theory on general scalar manifolds:
Symplectic covariance, gaugings and the momentum map,” J. Geom. Phys. 23, 111
(1997) hep-th/9605032.
[18] P. Aspinwall, private communication.
[19] B. de Wit and A. Van Proeyen, “Isometries of special manifolds,” hep-th/9505097
[20] M. Gunaydin and M. Zagermann, “The gauging of five-dimensional, N = 2 Maxwell-
Einstein supergravity theories coupled to tensor multiplets,” Nucl. Phys. B572 (2000)
131, hep-th/9912027; “The vacua of 5d, N = 2 gauged Yang-Mills/Einstein/tensor
supergravity: Abelian case,” Phys. Rev. D62 (2000) 044028, hep-th/0002228.
[21] A. Ceresole and G. Dall’Agata, “General matter coupled N =2, D = 5 gauged super-
gravity,” hep-th/0004111.
[22] R. Kallosh and A. Linde, “Supersymmetry and the brane world,” JHEP 0002, 005
(2000),hep-th/0001071.
[23] A. Lukas, B. A. Ovrut, K. S. Stelle and D. Waldram, “Heterotic M-theory in five
dimensions,” Nucl. Phys. B552, 246 (1999), hep-th/9806051.
[24] K. Behrndt and M. Cvetic, “Gauging of N = 2 supergravity hypermultiplet and novel
renormalization group flows,”hep-th/0101007.
[25] M. Gunaydin, G. Sierra and P. K. Townsend, “Gauging The D = 5 Maxwell-Einstein
Supergravity Theories: More On Jordan Algebras”, Nucl. Phys. B253, 573 (1985).
[26] A. Chou, R. Kallosh, J. Rahmfeld, S. Rey, M. Shmakova and W. K. Wong, “Critical
points and phase transitions in 5d compactifications of M-theory,” Nucl. Phys. B508,
147 (1997), hep-th/9704142.
[27] A. Khavaev and N. P. Warner, “A class of N = 1 supersymmetric RG flows from
five-dimensional N = 8 supergravity,” hep-th/0009159.
[28] K. Behrndt and M. Cvetic, “Anti-de Sitter vacua of gauged supergravities with 8
supercharges,” Phys. Rev. D61, 101901 (2000), hep-th/0001159.
[29] S. Gukov, “Solitons, superpotentials and calibrations,” Nucl. Phys. B574, 169 (2000),
hep-th/9911011.
[30] K. Behrndt and S. Gukov, “Domain walls and superpotentials from M theory on
Calabi-Yau three-folds”, Nucl. Phys. B580, 225 (2000), hep-th/0001082.
[31] T. R. Taylor and C. Vafa, “RR flux on Calabi-Yau and partial supersymmetry break-
ing,” Phys. Lett. B474, 130 (2000), hep-th/9912152.
[32] G. Curio, A. Klemm, D. Luest, S. Theisen, “On the Vacuum Structure of Type II
String Compactifications on Calabi-Yau Spaces with H-Fluxes,” hep-th/0012213.
[33] B. de Wit, P. G. Lauwers and A. Van Proeyen, “Lagrangians Of N=2 Supergravity -
Matter Systems,” Nucl. Phys. B255, 569 (1985).
[34] J. Michelson, “Compactifications of type IIB strings to four dimensions with non-
trivial classical potential,” Nucl. Phys. B495, 127 (1997), hep-th/9610151.
33
[35] S. Ferrara, R. Kallosh and A. Strominger, “N=2 extremal black holes,” Phys. Rev.
D52 (1995) 5412, hep-th/9508072
[36] S. Ferrara and R. Kallosh, “Supersymmetry and attractors,” Phys. Rev. D54 (1996)
1514, hep-th/9602136
[37] M. Gutperle, M. Spalinski, “Supergravity Instantons for N=2 Hypermultiplets,”
hep-th/0010192.
[38] W. A. Sabra, “Black holes in N=2 supergravity theories and harmonic functions,”
Nucl.Phys. B510 (1998) 247, hep-th/9704147; “General BPS black holes in five di-
mensions,” Mod. Phys. Lett. A13, 239 (1998), hep-th/9708103.
[39] work in progress.
[40] R. Kallosh, A. Linde, M. Shmakova, ”Supersymmetric Multiple Basin Attractors”,
JHEP 9911 (1999) 010
[41] R. Kallosh, ”Multivalued Entropy of Supersymmetric Black Holes,” JHEP 0001
(2000) 001
[42] S. Ferrara, G. W. Gibbons, R. Kallosh, “Black Holes and Critical Points in Moduli
Space,” Nucl.Phys. B500 (1997) 75, hep-th/9702103
[43] A. Chamseddine, S. Ferrara, G. Gibbons, R. Kallosh, Phys.Rev. D55 (1997) 3647,
hep-th/9610155.
[44] K. Behrndt, A. H. Chamseddine and W. A. Sabra, “BPS black holes in N=2 five
dimensional ADS supergravity,” Phys. Lett. B422,(1998) 97, hep-th/9807187.
[45] R. Britto-Pacumio, J. Michelson, A. Strominger and A. Volovich, “Lectures on super-
conformal quantum mechanics and multi-black hole moduli spaces”, hep-th/9911066.
[46] S. Ferrara, G. Gibbons and R. Kallosh, “Black holes and critical points in moduli
space”, hep-th/9702103.
[47] F. Denef, “Supergravity flows and D-brane stability”, hep-th/0005049
[48] S. Wu, “On the Instanton Complex of Holomorphic Morse Theory,” math.AG/9806118.
[49] M. Gunaydin, M. Zagermann, “The Gauging of Five-dimensional, N=2 Maxwell-
Einstein Supergravity Theories coupled to Tensor Multiplets,” Nucl.Phys. B572
(2000) 131.
[50] B. R. Greene, K. Schalm and G. Shiu, “Dynamical topology change in M theory,”
hep-th/0010207.
[51] J. Bagger and E. Witten, Nucl.Phys. B222 (1983) 1.
[52] Y.-K. E. Cheung, O.J. Ganor, M. Krogh, “On the Twisted (2,0) and Little String
Theories,” Nucl. Phys. B536 (1998) 175, hep-th/9805045.
[53] E. Bergshoeff, R. Kallosh, A. Van Proeyen, ”Supersymmetry in singular spaces,”
JHEP 0010 (2000) 033, hep-th/0007044.
34